CN110244734A - A kind of automatic driving vehicle paths planning method based on depth convolutional neural networks - Google Patents
A kind of automatic driving vehicle paths planning method based on depth convolutional neural networks Download PDFInfo
- Publication number
- CN110244734A CN110244734A CN201910537330.7A CN201910537330A CN110244734A CN 110244734 A CN110244734 A CN 110244734A CN 201910537330 A CN201910537330 A CN 201910537330A CN 110244734 A CN110244734 A CN 110244734A
- Authority
- CN
- China
- Prior art keywords
- path
- convolutional neural
- neural networks
- depth convolutional
- sampling
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000013439 planning Methods 0.000 title claims abstract description 59
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 43
- 238000005070 sampling Methods 0.000 claims abstract description 63
- 230000004888 barrier function Effects 0.000 claims abstract description 33
- 238000012549 training Methods 0.000 claims abstract description 27
- 238000013528 artificial neural network Methods 0.000 claims abstract description 20
- 238000012360 testing method Methods 0.000 claims abstract description 20
- 230000008569 process Effects 0.000 claims description 9
- 238000013461 design Methods 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- 230000006872 improvement Effects 0.000 claims description 3
- 238000003062 neural network model Methods 0.000 claims description 3
- 230000000153 supplemental effect Effects 0.000 claims description 3
- 238000011478 gradient descent method Methods 0.000 claims description 2
- 241000208340 Araliaceae Species 0.000 claims 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 claims 1
- 235000003140 Panax quinquefolius Nutrition 0.000 claims 1
- 235000008434 ginseng Nutrition 0.000 claims 1
- 210000002569 neuron Anatomy 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 239000013598 vector Substances 0.000 description 4
- 239000012141 concentrate Substances 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 210000005036 nerve Anatomy 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000005267 amalgamation Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000004218 nerve net Anatomy 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000006641 stabilisation Effects 0.000 description 1
- 238000011105 stabilization Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
- G05D1/0214—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory in accordance with safety or protection criteria, e.g. avoiding hazardous areas
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
- G05D1/0221—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving a learning process
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
- G05D1/0223—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving speed control of the vehicle
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0231—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means
- G05D1/0246—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using a video camera in combination with image processing means
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Aviation & Aerospace Engineering (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Electromagnetism (AREA)
- Traffic Control Systems (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of automatic driving vehicle paths planning methods based on depth convolutional neural networks；Depth convolutional neural networks model is constructed first；Then barrier grid map and corresponding reference path of the acquisition vehicle under different driving environments, for constructing depth convolutional neural networks model training, verifying and the sample data library of test；Finally, the depth convolutional neural networks model that training and test are completed is used for automatic driving vehicle, input barrier grid map and reference path in real time to neural network, generate crucial sampling area, and the path planning algorithm based on sampling is utilized, sampling and plane-generating path from crucial sampling area.The present invention overcomes in end-to-end method, prediction error can not overcome the problems, such as, keep the driving trace of automatic driving vehicle more stable and controllable；The shortcomings that selecting Path Method to lack flexibility and adaptability from preset path set is also overcomed, improves the flexibility of path planning, while and ensure that path quality.
Description
Technical field
The present invention relates to depth convolutional neural networks and automatic Pilot fields, are rolled up more particularly, to one kind based on depth
The automatic driving vehicle paths planning method of product neural network.
Background technique
In recent years, unmanned technology is increasingly becoming the research hotspot of domestic and international colleges and universities and enterprise, also because it is commercially being answered
Cause public concern with exposing in test again and again.Unmanned vehicle is during realizing automatic Pilot, the road of autonomous driving system
Path planning subsystem is the key foundation of unmanned vehicle automatic Pilot, is to guarantee unmanned vehicle safety and stability, collisionless normally travel
Premise.
Depth convolutional neural networks are multiple in image classification, target detection, scene cut etc. with its superior performance
Field obtains immense success.Compared with traditional method, depth convolutional neural networks can be realized end-to-end training, without by hand
Design feature can learn automatically, and accuracy rate is higher, robustness is stronger.Just because of depth convolutional Neural network the advantages of, certainly
Also it is more and more applied in the dynamic path planning field driven.There are many researchs to concentrate on realization planning path end to end
It generates, i.e., will input network to the picture at visual angle before unmanned vehicle, directly generate effective path.Also some research concentrates on
The control instruction that unmanned vehicle is directly exported using the preceding image to visual angle is merged path planning and motion control.There are also some
Research concentrates on the key message needed for planning using depth convolutional Neural network learning path, accelerates road using the information of extraction
The quality of diameter planning speed and rectifying plan path.
Application No. is 108469815 A of CN, discloses a kind of computer deep learning self-navigation based on intention and drive
Method is sailed, traditional High Accuracy Radar map is not needed, common digital map navigation result can be used to automatic Pilot.Road
Diameter planning is served only for being intended to extract, and the intention is extracted, for automatic Pilot, it is intended that is extracted and is derived from path planning.The bottom
Decision of the layer based on intention, learns various driving efficiencies using depth learning by imitation.Application No. is 108875998 A of CN, relate to
And a kind of automatic driving vehicle method and system for planning, global path planning, local behavior planning and movement are advised from system
It draws and is merged, avoid the isolated distribution of each module and lack logical level；From information flow using from top to down with to it is lower and on
Amalgamation mode improves system robustness.But in the above scheme, first, it extracts and is intended to from global path, then using deep
Convolutional network is spent based on the control for being intended to progress vehicle, and being equivalent to realizes path planning and control using depth learning by imitation
The fusion of system is that method, the method can not overcome control of the prediction error of neural network to vehicle to one kind end to end
Influence.Moreover, training process needs great amount of samples, meanwhile, it is difficult to predict safety and stabilities for the driving trace of unmanned vehicle
It is insufficient；Second, path planning task, this side are completed by selecting the path for being suitble to current environment from preset path set
Method is due to having preset optional path, it is more difficult to adapt to traffic environment complicated and changeable.Meanwhile road is not carried out for scene
The optimization of diameter, the quality in the path of generation are relatively difficult to guarantee.
Summary of the invention
The present invention is the control for overcoming the above-mentioned prediction error that can not overcome neural network in the prior art to vehicle
Influence defect, a kind of automatic driving vehicle paths planning method based on depth convolutional neural networks is provided, depth is utilized
Convolutional neural networks automatically extract feature from reference path and barrier grid map and obtain crucial sampling area, then utilize
Path planning algorithm based on sampling samples from crucial sampling area and generates path, thus overcome in end-to-end method,
Prediction error can not overcome the problems, such as, keep the driving trace of automatic driving vehicle more stable and controllable.
In order to solve the above technical problems, the technical solution adopted by the present invention is that: it is a kind of based on depth convolutional neural networks
Automatic driving vehicle paths planning method, comprising the following steps:
S1. depth convolutional neural networks model is constructed, for generating crucial sampling area；
S2. barrier grid map and corresponding reference path of the acquisition vehicle under different driving environments, using having
Targetedly path planning algorithm is cooked up according to barrier grid map and corresponding reference path in this barrier grid map
With the optimal path under the conditions of reference path；The point on the optimal path is taken, as this barrier grid map and reference path institute
Corresponding key sampling area, that is, input corresponding label；It is established later for the training of depth convolutional neural networks, verifying and is surveyed
The sample data library of examination, each training examples includes barrier grid map, reference path and crucial sampling area in database；
S3. training, verifying and test depth convolutional neural networks；The database established in step S2 is divided into three parts:
Training set, verifying collection and test set；Neural network is trained with the sample in training set first, most using stochastic gradient descent method
Smallization loss function is trained, until loss restrains；Then the neural network that training is completed is tested with test set；
S4. the neural network constructed in the step S1 of test completion is used for path planning；Barrier grid is inputted first
Figure and reference path generate crucial sampling area using neural network；Then use the paths planning method based on sampling from pass
It is sampled in key sampling area and generates path.
Paths planning method proposed by the present invention based on barrier grid map and reference path stabilization and can be quickly generated
High quality path meets automatic driving vehicle path planning (part) mission requirements, realizes that automatic driving vehicle cruises safely.This
Invention constructs depth convolutional neural networks model first, for generating crucial sampling area；Then acquisition vehicle is in different driving
Barrier grid map and corresponding reference path under environment, for constructing depth convolutional neural networks model training, verifying
With the sample data library of test；Finally, the depth convolutional neural networks model that training and test are completed is used for automatic Pilot vehicle
, it inputs barrier grid map and reference path in real time to neural network, generates crucial sampling area, and using based on sampling
Path planning algorithm, sampling and plane-generating path from crucial sampling area.
The present invention is automatically extracted feature using depth convolutional neural networks and is obtained from reference path and barrier grid map
To crucial sampling area, the path planning algorithm based on sampling is then utilized, path is sampled and generated from crucial sampling area,
To overcome in end-to-end method, prediction error can not overcome the problems, such as, keep the driving trace of automatic driving vehicle more stable
With it is controllable.Meanwhile the shortcomings that selecting Path Method to lack flexibility and adaptability from preset path set is also overcomed, it utilizes
Depth convolutional neural networks generate targetedly crucial sampling area for different driving environments, and utilize the path based on sampling
Planning algorithm targetedly carries out coordinates measurement and path quality optimization, improves the flexibility of path planning, while and ensure that
Path quality.
The present invention is based on the path plannings that depth convolutional neural networks carry out automatic driving vehicle, roll up compared to based on depth
Method, the present invention can preferably deal with the prediction error bring vehicle movement of neural network to product neural network end to end
Unstability, robustness are stronger.Compared to the paths planning method of fixed route collection, it is more that the present invention can preferably deal with complexity
The driving environment of change, while the path quality planned is more stable.
Further, the neural network model constructed in the S1 step includes main split and secondary branch；Main split
Input is barrier grid map (picture), and main includes 1 depth convolution sub-network for being used for image characteristics extraction, such as
ResNet50, VGG and Inception etc., 1 convolutional layer, 1 full articulamentum and several supplemental characteristic process layers；Secondary branch
Input is reference path (two-dimensional matrix), mainly includes 1 full articulamentum；The output of main split and secondary branch can be spliced
Come, and passes through 1 full articulamentum, final output key sampling area.
Further, the reference path in S2 step is defined as orderly point set: RP={ p0,p1,…,pn},pi∈SE(2),
n∈Z+；Wherein to the element p in any RPi, piIn global path, it is located in front of automatic driving vehicle, apart from vehicle geometry
Euler's distance at center is ie, and e is the unit distance of a setting；Meanwhile piCoordinate origin is the origin of vehicle axis system；For side
Just Processing with Neural Network, barrier grid map are converted into RGB figure, and the place coordinate system of barrier grid map is also vehicle coordinate
System；
Further；Crucial sampling area in S2 step is defined as unordered point set: KR={ k0,k1,…,kn},ri∈SE
(2),m∈Z+, using vehicle axis system；To generate crucial sampling area, the path planning based on sampling by design is utilized
Algorithm, as RRT* method biases the method for sampling using Gauss and constantly sample, directly using the point in reference path as sample reference point
Until the path of generation no longer increases with sampling number and had clear improvement, the point on this optimal path is then taken to constitute this barrier
Hinder corresponding crucial sampling area KR under the conditions of object grid map and reference path.
Further, it in the S3 step, in each training process, is verified with verifying collection, to check whether
Over-fitting；In the training stage, use mean value absolute deviation as loss function；In test phase, when to prediction KRpMiddle any pointItself and label KR1Middle corresponding pointsBetween Euler's distanceWhen all setting up, it is believed that prediction is correct, otherwise pre-
Sniffing misses.
Further, in the S4 step, using the paths planning method based on sampling from crucial sampling area KR
When sampling, using the point in KR as sample-offset center, using Gauss biasing means, it is biased sampling.This sampling process is similar
When generating crucial sampling area label in step S2, the process that is sampled from reference path.Difference is that the former adopts from key
Sample samples in region, and the latter samples from reference path.
Compared with prior art, beneficial effect is:
1. the present invention is based on depth convolutional neural networks to generate crucial sampling area, rather than directly generates planning path, drop
The degree of difficulty of low network design and training；
2. the present invention extracts the feature in barrier grid map using depth convolutional neural networks, have more path planning
Targetedly, scene complicated and changeable can preferably be dealt with；
3. the present invention is sampled from the crucial sampling area that neural network generates using the path planning algorithm based on sampling
And path is generated, the method can effectively overcome the unstability of neural network prediction error bring path planning, make system
Robustness is stronger.Also make the quality in the path of planning more stable, will not occur stronger fluctuation because of environmental change.
Detailed description of the invention
Fig. 1 is the paths planning method overall flow figure the present invention is based on depth convolutional neural networks.
Fig. 2 is the model structure for the depth convolutional neural networks that the present invention constructs.
Specific embodiment
Attached drawing only for illustration, is not considered as limiting the invention；In order to better illustrate this embodiment, attached
Scheme certain components to have omission, zoom in or out, does not represent the size of actual product；To those skilled in the art,
The omitting of some known structures and their instructions in the attached drawings are understandable.Being given for example only property of positional relationship is described in attached drawing
Illustrate, is not considered as limiting the invention.
As shown in Figure 1, a kind of automatic driving vehicle paths planning method based on depth convolutional neural networks, including it is following
Step:
Step 1. constructs depth convolutional neural networks model, and the neural network model of building includes main split and secondary branch；
The input of main split is barrier grid map (picture), and main includes 1 depth convolution sub-network for being used for image characteristics extraction,
Such as ResNet50, VGG and Inception, 1 convolutional layer, 1 full articulamentum and several supplemental characteristic process layers；Secondary branch
Input be reference path (two-dimensional matrix), mainly include 1 full articulamentum；The output of main split and secondary branch can be spliced
Come, and passes through 1 full articulamentum, final output key sampling area.
Step 2. collects different barrier grid maps, reference path.And with the path planning algorithm based on sampling, RRT*
Algorithm is based on barrier grid map and reference path, generates corresponding optimal path, takes the point on optimal path to constitute key and adopts
Sample region, as label corresponding with this barrier grid map and reference path.Sample data library is established by this step.
The sample data library that step 3. establishes step 2 is divided into training set, verifying collection and test set, and sample quantity ratio is about
For 7:1:2.First with training set training depth convolutional neural networks, each step trained all is needed with verifying collection verifying nerve net
The effect of network.Then the prediction of test neural network is gone to show with test set.
Step 4. is using step 3 training and tests the depth convolutional neural networks of completion according to the barrier grid inputted in real time
Trrellis diagram and reference path predict crucial sampling area.
The crucial sampling area that step 4 prediction obtains is integrated into the path planning algorithm RRT* based on sampling by step 5.
In.The RRT* algorithm samples from crucial sampling area and ultimately generates the path planned by automatic driving vehicle.
As shown in Fig. 2, the model structure of the depth convolutional neural networks constructed for the present invention；Wherein, in figure
Resnet50 represents Resnet50 convolutional neural networks；1x1conv represents convolution kernel as the convolution operation of 1x1x256 in figure；Figure
One layer of neuron is launched into one-dimensional by middle Flatten representative；BN batch normalization operations in figure；FC is represented one layer of nerve in figure
It is first to be connected entirely with another layer of neuron.Concat, which is represented, in figure merges one layer of one-dimensional nerve of layer for two layers of one-dimensional neuron
Member.Dropout represents Dropout operation in figure；Reshape represents the dimension of one layer of neuron of reconstruct in figure.Concrete operations step
Suddenly include:
1. primary input be barrier grid map, be (500,500,3) RGB picture, successively by Resnet50, (1,1,
256) after the convolutional layer of convolution kernel, Flatten layers, BN layers and FC layers, it is processed into the feature vector of 256 dimensions；
It is that the matrix of (4,3) is extended to successively after flatten layers and FC layers 2. auxiliary input is reference path
The feature vector of 256 dimensions；
3. the feature vector of step 1 and 2 two 256 dimensions is merged into 512 dimensional feature vectors, successively pass through BN layers, FC
After layer, Dropout layers, FC layers and Reshape layers, the output of neural network is obtained, i.e., crucial sampling area, is the square of (4,3)
Battle array.
The present invention is directed to the path planning task of automatic driving vehicle, generates crucial sampling based on depth convolutional neural networks
Region, and the crucial sampling area generated is combined, path is generated using the path planning algorithm based on sampling, can be improved automatic
Drive the speed of vehicle path planning, the stability of path quality and the robustness of path planning subsystem.
Obviously, the above embodiment of the present invention be only to clearly illustrate example of the present invention, and not be pair
The restriction of embodiments of the present invention.For those of ordinary skill in the art, may be used also on the basis of the above description
To make other variations or changes in different ways.There is no necessity and possibility to exhaust all the enbodiments.It is all this
Made any modifications, equivalent replacements, and improvements etc., should be included in the claims in the present invention within the spirit and principle of invention
Protection scope within.
Claims (6)
1. a kind of automatic driving vehicle paths planning method based on depth convolutional neural networks, which is characterized in that including following
Step:
S1. depth convolutional neural networks model is constructed；
S2. barrier grid map and corresponding reference path of the acquisition vehicle under different driving environments, is directed to using having
The path planning algorithm of property is cooked up according to barrier grid map and corresponding reference path in this barrier grid map and ginseng
Examine the optimal path under path condition；The point on the optimal path is taken, as corresponding to this barrier grid map and reference path
Crucial sampling area, that is, input corresponding label；It is established later for the training of depth convolutional neural networks, verifying and test
Sample data library, each training examples includes barrier grid map, reference path and crucial sampling area in database；
S3. training, verifying and test depth convolutional neural networks；The database established in step S2 is divided into three parts: training
Collection, verifying collection and test set；First with the sample training neural network in training set, minimized using stochastic gradient descent method
Loss function is trained, until loss restrains；Then the neural network that training is completed is tested with test set；
S4. the neural network constructed in the step S1 of test completion is used for path planning；First input barrier grid map and
Reference path generates crucial sampling area using neural network；Then the paths planning method based on sampling is used to adopt from key
It is sampled in sample region and generates path.
2. the automatic driving vehicle paths planning method according to claim 1 based on depth convolutional neural networks, special
Sign is that the neural network model constructed in the S1 step includes main split and secondary branch；The input of main split is obstacle
Object grid map, it is main include 1 for the depth convolution sub-network of image characteristics extraction, 1 convolutional layer, 1 full articulamentum and
Several supplemental characteristic process layers；The input of secondary branch is reference path, mainly includes 1 full articulamentum；Main split and secondary branch
Output can be spliced, and by 1 full articulamentum, final output key sampling area.
3. the automatic driving vehicle paths planning method according to claim 2 based on depth convolutional neural networks, special
Sign is that the reference path in the S2 step is defined as orderly point set: RP={ p0, p1..., pn, pi∈ SE (2), n ∈
Z+；Wherein to the element p in any RPi, piIn global path, it is located in front of automatic driving vehicle, apart from vehicle geometric center
Euler distance be ie, e be one setting unit distance；Meanwhile piCoordinate origin is the origin of vehicle axis system；For convenience of mind
Through network processes, barrier grid map is converted into RGB figure, and the place coordinate system of barrier grid map is also vehicle axis system.
4. the automatic driving vehicle paths planning method according to claim 3 based on depth convolutional neural networks, special
Sign is that the crucial sampling area in the S2 step is defined as unordered point set: KR={ k0, k1..., kn, ri∈SE
(2), m ∈ Z+, using vehicle axis system；To generate crucial sampling area, the path planning based on sampling by design is utilized
Algorithm biases the method for sampling using Gauss and constantly samples using the point in reference path as sample reference point, until the path generated
No longer with sampling number increase and until having clear improvement, then take the point on this optimal path constitute this barrier grid map and
Corresponding key sampling area KR under the conditions of reference path.
5. the automatic driving vehicle paths planning method according to claim 4 based on depth convolutional neural networks, special
Sign is, in the S3 step, in each training process, is verified with verifying collection, to check whether over-fitting；It is instructing
Practice the stage, uses mean value absolute deviation as loss function；In test phase, when to prediction KRpMiddle any pointItself and label
KRlMiddle corresponding pointsBetween Euler's distanceWhen all setting up, it is believed that prediction is correct, otherwise prediction error.
6. the automatic driving vehicle paths planning method according to claim 4 based on depth convolutional neural networks, special
Sign is, in the S4 step, when being sampled from crucial sampling area KR using the paths planning method based on sampling, with KR
In point be that sample-offset center using Gauss biasing means is biased sampling.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910537330.7A CN110244734B (en) | 2019-06-20 | 2019-06-20 | Automatic driving vehicle path planning method based on deep convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910537330.7A CN110244734B (en) | 2019-06-20 | 2019-06-20 | Automatic driving vehicle path planning method based on deep convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110244734A true CN110244734A (en) | 2019-09-17 |
| CN110244734B CN110244734B (en) | 2021-02-05 |
Family
ID=67888516
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910537330.7A Active CN110244734B (en) | 2019-06-20 | 2019-06-20 | Automatic driving vehicle path planning method based on deep convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110244734B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110609560A (en) * | 2019-10-29 | 2019-12-24 | 广州高新兴机器人有限公司 | Mobile robot obstacle avoidance planning method and computer storage medium |
| CN110986949A (en) * | 2019-12-04 | 2020-04-10 | 日照职业技术学院 | Path identification method based on artificial intelligence platform |
| CN111002980A (en) * | 2019-12-10 | 2020-04-14 | 苏州智加科技有限公司 | Road obstacle trajectory prediction method and system based on deep learning |
| CN111289003A (en) * | 2020-02-06 | 2020-06-16 | 广州小马智行科技有限公司 | Path planning method, device, system, storage medium and processor |
| CN111723632A (en) * | 2019-11-08 | 2020-09-29 | 珠海达伽马科技有限公司 | Ship tracking method and system based on twin network |
| CN113296500A (en) * | 2021-04-30 | 2021-08-24 | 浙江吉利控股集团有限公司 | Local path planning method and system |
| CN113311828A (en) * | 2021-05-08 | 2021-08-27 | 武汉理工大学 | Unmanned vehicle local path planning method, device, equipment and storage medium |
| CN113359717A (en) * | 2021-05-26 | 2021-09-07 | 浙江工业大学 | Mobile robot navigation obstacle avoidance method based on deep reinforcement learning |
| CN115617054A (en) * | 2021-07-15 | 2023-01-17 | 中移***集成有限公司 | Path planning method and system, electronic device and readable storage medium |
| WO2023201952A1 (en) * | 2022-04-21 | 2023-10-26 | 合众新能源汽车股份有限公司 | Method and apparatus for determining optimal traveling trajectory of vehicle |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106767860A (en) * | 2016-11-21 | 2017-05-31 | 江苏大学 | A kind of method that intelligent automobile path planning search time is shortened based on heuristic search algorithm |
| CN107742304A (en) * | 2017-10-10 | 2018-02-27 | 广州视源电子科技股份有限公司 | Method and device for determining movement track, mobile robot and storage medium |
| CN107817798A (en) * | 2017-10-30 | 2018-03-20 | 洛阳中科龙网创新科技有限公司 | A kind of farm machinery barrier-avoiding method based on deep learning system |
| CN107862293A (en) * | 2017-09-14 | 2018-03-30 | 北京航空航天大学 | Radar based on confrontation generation network generates colored semantic image system and method |
| CN108415032A (en) * | 2018-03-05 | 2018-08-17 | 中山大学 | A kind of point cloud semanteme map constructing method based on deep learning and laser radar |
| US20180292835A1 (en) * | 2016-12-09 | 2018-10-11 | Zendrive, Inc. | Method and system for risk modeling in autonomous vehicles |
| CN108875998A (en) * | 2018-04-20 | 2018-11-23 | 北京智行者科技有限公司 | A kind of automatic driving vehicle method and system for planning |
| CN108985194A (en) * | 2018-06-29 | 2018-12-11 | 华南理工大学 | A kind of intelligent vehicle based on image, semantic segmentation can travel the recognition methods in region |
| CN109242003A (en) * | 2018-08-13 | 2019-01-18 | 浙江零跑科技有限公司 | Method is determined based on the vehicle-mounted vision system displacement of depth convolutional neural networks |
| CN109263639A (en) * | 2018-08-24 | 2019-01-25 | 武汉理工大学 | Driving path planing method based on state Grid Method |
| CN109557928A (en) * | 2019-01-17 | 2019-04-02 | 湖北亿咖通科技有限公司 | Automatic driving vehicle paths planning method based on map vector and grating map |
| CN109858372A (en) * | 2018-12-29 | 2019-06-07 | 浙江零跑科技有限公司 | A kind of lane class precision automatic Pilot structured data analysis method |
| CN109901595A (en) * | 2019-04-16 | 2019-06-18 | 山东大学 | A kind of automated driving system and method based on monocular cam and raspberry pie |
-
2019
- 2019-06-20 CN CN201910537330.7A patent/CN110244734B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106767860A (en) * | 2016-11-21 | 2017-05-31 | 江苏大学 | A kind of method that intelligent automobile path planning search time is shortened based on heuristic search algorithm |
| US20180292835A1 (en) * | 2016-12-09 | 2018-10-11 | Zendrive, Inc. | Method and system for risk modeling in autonomous vehicles |
| CN107862293A (en) * | 2017-09-14 | 2018-03-30 | 北京航空航天大学 | Radar based on confrontation generation network generates colored semantic image system and method |
| CN107742304A (en) * | 2017-10-10 | 2018-02-27 | 广州视源电子科技股份有限公司 | Method and device for determining movement track, mobile robot and storage medium |
| CN107817798A (en) * | 2017-10-30 | 2018-03-20 | 洛阳中科龙网创新科技有限公司 | A kind of farm machinery barrier-avoiding method based on deep learning system |
| CN108415032A (en) * | 2018-03-05 | 2018-08-17 | 中山大学 | A kind of point cloud semanteme map constructing method based on deep learning and laser radar |
| CN108875998A (en) * | 2018-04-20 | 2018-11-23 | 北京智行者科技有限公司 | A kind of automatic driving vehicle method and system for planning |
| CN108985194A (en) * | 2018-06-29 | 2018-12-11 | 华南理工大学 | A kind of intelligent vehicle based on image, semantic segmentation can travel the recognition methods in region |
| CN109242003A (en) * | 2018-08-13 | 2019-01-18 | 浙江零跑科技有限公司 | Method is determined based on the vehicle-mounted vision system displacement of depth convolutional neural networks |
| CN109263639A (en) * | 2018-08-24 | 2019-01-25 | 武汉理工大学 | Driving path planing method based on state Grid Method |
| CN109858372A (en) * | 2018-12-29 | 2019-06-07 | 浙江零跑科技有限公司 | A kind of lane class precision automatic Pilot structured data analysis method |
| CN109557928A (en) * | 2019-01-17 | 2019-04-02 | 湖北亿咖通科技有限公司 | Automatic driving vehicle paths planning method based on map vector and grating map |
| CN109901595A (en) * | 2019-04-16 | 2019-06-18 | 山东大学 | A kind of automated driving system and method based on monocular cam and raspberry pie |
Non-Patent Citations (3)
| Title |
|---|
| RAMAZAN YENICERI等: "A New CNN Based Path Planning Algorithm Improved by the Doppler Effect", 《 DEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING 》 * |
| 唐静: "基于卷积神经网络的道路场景感知算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 郭笑笑等: "基于无人驾驶汽车的避障算法综述", 《计算机科学》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110609560A (en) * | 2019-10-29 | 2019-12-24 | 广州高新兴机器人有限公司 | Mobile robot obstacle avoidance planning method and computer storage medium |
| CN111723632A (en) * | 2019-11-08 | 2020-09-29 | 珠海达伽马科技有限公司 | Ship tracking method and system based on twin network |
| CN111723632B (en) * | 2019-11-08 | 2023-09-15 | 珠海达伽马科技有限公司 | Ship tracking method and system based on twin network |
| CN110986949A (en) * | 2019-12-04 | 2020-04-10 | 日照职业技术学院 | Path identification method based on artificial intelligence platform |
| CN111002980A (en) * | 2019-12-10 | 2020-04-14 | 苏州智加科技有限公司 | Road obstacle trajectory prediction method and system based on deep learning |
| CN111289003A (en) * | 2020-02-06 | 2020-06-16 | 广州小马智行科技有限公司 | Path planning method, device, system, storage medium and processor |
| CN113296500A (en) * | 2021-04-30 | 2021-08-24 | 浙江吉利控股集团有限公司 | Local path planning method and system |
| CN113311828A (en) * | 2021-05-08 | 2021-08-27 | 武汉理工大学 | Unmanned vehicle local path planning method, device, equipment and storage medium |
| CN113311828B (en) * | 2021-05-08 | 2023-07-25 | 武汉理工大学 | Unmanned vehicle local path planning method, device, equipment and storage medium |
| CN113359717A (en) * | 2021-05-26 | 2021-09-07 | 浙江工业大学 | Mobile robot navigation obstacle avoidance method based on deep reinforcement learning |
| CN115617054A (en) * | 2021-07-15 | 2023-01-17 | 中移***集成有限公司 | Path planning method and system, electronic device and readable storage medium |
| WO2023201952A1 (en) * | 2022-04-21 | 2023-10-26 | 合众新能源汽车股份有限公司 | Method and apparatus for determining optimal traveling trajectory of vehicle |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110244734B (en) | 2021-02-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110244734A (en) | A kind of automatic driving vehicle paths planning method based on depth convolutional neural networks | |
| CN110956651B (en) | Terrain semantic perception method based on fusion of vision and vibrotactile sense | |
| KR102296507B1 (en) | Method for tracking object by using convolutional neural network including tracking network and computing device using the same | |
| CN111862126B (en) | Non-cooperative target relative pose estimation method combining deep learning and geometric algorithm | |
| CN108415032B (en) | Point cloud semantic map construction method based on deep learning and laser radar | |
| EP3686779B1 (en) | Method and device for attention-based lane detection without post-processing by using lane mask and testing method and testing device using the same | |
| Li et al. | Learning object-wise semantic representation for detection in remote sensing imagery | |
| CN105930819A (en) | System for real-time identifying urban traffic lights based on single eye vision and GPS integrated navigation system | |
| Bojarski et al. | The NVIDIA pilotnet experiments | |
| CN102915039A (en) | Multi-robot combined target searching method of animal-simulated space cognition | |
| CN110033411A (en) | The efficient joining method of highway construction scene panoramic picture based on unmanned plane | |
| CN109948707A (en) | Model training method, device, terminal and storage medium | |
| Pham et al. | Road damage detection and classification with YOLOv7 | |
| CN113255589B (en) | Target detection method and system based on multi-convolution fusion network | |
| CN112464912B (en) | Robot end face detection method based on YOLO-RGGNet | |
| EP3690397B1 (en) | Method and device for localization of autonomous vehicle for route planning by using attention-driven landmark detection | |
| JP2020119523A (en) | Method for detecting pseudo-3d bounding box and device using the same | |
| CN110109465A (en) | A kind of self-aiming vehicle and the map constructing method based on self-aiming vehicle | |
| JP6810432B2 (en) | A method of detecting a pseudo 3D bounding box used for military purposes, smartphones or virtual driving on a CNN platform that can switch modes according to the conditions of the object, and a device using this | |
| CN104680167A (en) | Aurora oval position determining method based on deep learning | |
| CN111507161A (en) | Method and apparatus | |
| CN112833891A (en) | Road data and lane-level map data fusion method based on satellite film recognition | |
| CN114119757A (en) | Image processing method, apparatus, device, medium, and computer program product | |
| Vasić et al. | Deep semantic image segmentation for UAV-UGV cooperative path planning: A car park use case | |
| CN117671647B (en) | Multitasking road scene perception method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |